The well-known edge server bottleneck is shown in this paper to be the result of a mismatch between the general-purpose architecture and the special-purpose data functions it was not intended to perform exclusively. Six overheads are identified that contribute significantly to the bottleneck. To solve this bottleneck problem, a solution is to apply the principle of separation between control and data functions. While the idea is not new, the application of this principle to the edge server architecture is novel in the convergence of three technologies: network, server, and storage. Notable performance has been obtained with this approach using ASIC implementation for TCP or UDP streaming applications.